Metal-Catalysed Reactions of Hydrocarbons / 02-Small Metal Particles and Supported Metal Catalysts

structure, i.e., the phases present and the surface composition as a function of total composition.279 These considerations were briefly treated in Section 2.1.1, and theoretical methods applied to surface composition are mentioned in Section 2.5.5. Segregation of the component of lower surface energy is expected to occur, although it does not always with NiCu/SiO2;280 the minimum concentration of that component needed to secure complete coverage increases as the particle size falls. Colloidal ruthenium-platinum particles were examined by TEM and EXAFS, and were shown to retain the fcc structure at all compositions.281

2.5.3. Energetic Properties

Notwithstanding the shorter bond lengths usually found between surface atoms and those beneath them, implying, as they do, stronger bonds and a ‘skin’ effect analogous to the surface tension of liquids, such atoms experience larger vibration amplitudes and frequencies, and enhanced mobility relative to bulk atoms269,270,282 (Section 1.2.2). Melting temperature decreases with particle size, and for gold it may approach ambient temperature for the smallest particles.113,218,283,284 The activation energy for surface mobility is much less than the sublimation energy (typically 15%), so that even for metals of quite high melting temperature the surface layer or zone may be semi-fluid (or quasi-molten) at high temperatures. The regions in which such a phase and other metastable phase can exist are shown schematically as a phase diagram in Figure 2.7. If the surface of a particle has passed through a semi-fluid stage (for example, during reduction), it may appear to be amorphous when examined at room temperature. These observations form an additional cogent reason for mistrusting arguments about particle size effects that are based solely on geometric considerations.

The excess free energy associated with small platinum particles can be measured by the potential at which they are oxidised to Pt4+ in the presence of chloride ion.285 An auxiliary redox system (Fe3+/Fe2+) had to be used, and the platinum particles then acted as a microelectrode, taking the reversible potential of the ferrous-ferric equilibrium, which was calculated by the Nernst equation. Use of different concentrations of ferric ion allowed the potential at which platinum atoms were oxidised to be determined, and from the equation

bulk metal was determined. For alumina-supported particles of mean size 21 nm this was found to be −54 kJ mol−1 and for 3.7 nm particles having a broad size distribution –108 kJ mol−1, i.e. small particles were more easily oxidised than large ones.

2.5.4. Electronic Properties209,286

Since the role of the geometric factor in the behaviour of small metal particles may be obscured by the effects described above, we may expect the electronic factor to assume a larger importance. Qualitatively it is apparent that lowering particle size must mean progressive loss of metallic character as gauged by commonality of the valence electrons, that is to say, by their bandwidth; ultimately the band structure relapses to an energy level structure when the particle contains only a few atoms (Section 2.5.1). This argument is the reverse of that used to explain why and how bands appear when particle size in increased.3 To a first approximation the spacing δ between energy levels is given by

where E F is the Fermi level energy224 and n the number of atoms in the particle. When separation between adjacent levels is greater than the thermal energy kT, they begin to act individually, and metallic character starts to disappear. For a value of E F of 10 eV (a typical value), the critical size at room temperature is about 400 atoms: but note that at the borderline raising the temperature should result in an increase in metal like behaviour. By some criteria full bulk metallic character may start to be seen in particles which have no more than 150–200 atoms;6 at this size about 60–70 % of the atoms are still on the ‘surface’, but only about 20–25 % of valences are not used in bonding atoms to each other (Figure 2.5).

Changes in band structure with particle size may be followed by ultraviolet photoelectron spectroscopy (UPS). They can be summarised by saying that (i) the density of states at the Fermi energy E F increases, (ii) the band becomes less sharp and develops features at lower energy, and (iii) the centre of ‘gravity’ of the band moves to lower energies, as particle size increases. These are the expected consequences of the development of the collective behaviour of electrons, and are supported by measurements of valence-electron binding energies by X-ray photoelectron spectroscopy (XPS). These decrease with growing particle size,276,287 but the effects are so small (usually 1–2 eV), and it was initially unclear whether they were caused by initialor final-state effects. The former seems now to be the generally accepted explanation.218

The energy quantum used in NMR is much smaller than that needed for electron spectroscopy,288,289 but the response is less sensitive, and a large sample is needed; also the interpretation is even less straightforward. Supported metal catalysts are very suitable for study and the 195Pt nucleus has been extensively examined (Section 2.4.2). This dependence of NMR amplitude on field/frequency shows separate, if not well resolved, peaks at 1.138 and 1.10 G kHz−1, corresponding respectively to Knight shifts of –3.34 and zero percent and thus to bulk and surface atoms. The plots have been imaginatively deconvoluted, but no use

has been made of the individual peak areas. The results have been manipulated to yield estimates of the local density of states (LDOS) at the Fermi surface224 (see the next section); this is much less for surface atoms than for those below, as would be expected from the UPS and XPS results quoted above. It is also stated to be independent of the type of support (SiO2, TiO2, Al2O3) and method of preparation (impregnation, ion exchange, colloidal process), and to depend solely on dispersion.

There have been a number of studies on the particle size dependence of ferromagnetic behaviour, but fewer on paramagnetic. Small palladium particles have however been shown to have lower paramagnetic susceptibility than large particles.168

Heats of adsorption of hydrogen on, and of dissolution in, supported palladium particles were size-independent down to 3 nm, but thereafter decreased significantly;290 those of oxygen on rhodium, palladium and platinum fell with increasing size.291

2.5.5. Theoretical Methods

Small metal particles have been a happy hunting ground for theoreticians. By performing calculations by a variety of theoretical procedures (LCAO-MO Xα scattered wave, density functional theory with various approximations), it has been possible to explore two main points: (i) the relative stabilities of structures of fivefold symmetry and those having the normal bulk form, and (ii) the emergence of electron energy bands with increasing particle size. The conclusions of this body of work have been summarised, and it is not surprising to find that experimental observations find a good measure of support from them. Indeed the cynic might say that a theory could be produced to account for any experimental result. Nevertheless it is a welcome check on the validity of computational procedures that they are able to confirm so closely what is in fact seen.

Particular attention has also been given to deriving the local density of states at the Fermi level (E F -LDOS).224,288 Calculations performed using the tight binding approximation confirm the result noted above, that this quantity decreases as all particles become smaller, i.e. with decreasing mean CN. These calculations have been pursued to their logical limit, by assigning values of E F -LDOS to atoms of each individual CN at the surface of a complete cubo-octahedron: they also decrease with CN. The chemically minded reader, if worried by the use of Band Theory language to describe the electronic structure of small particles, can change LDOS to a statement of the energies and occupancies of the valence orbitals.

Although as we shall see there is much experimental work that bears on the particle size dependence of catalytic behaviour, its interpretation is speculative, and there is little that informs directly on how surface atoms differ from

those inside. Perhaps the most immediate evidence comes from examining the

vibration frequency in the carbon monoxide molecule decreases with particle size on Ir/Al2O3292 and Pt/SiO2293 the which may be due to increasing free-atom like

character of surface atoms, leading to enhanced electron density and back donation into π * antibonding orbitals.294 Interpretation of the spectra of chemisorbed carbon monoxide can also lead to estimates of the relative areas of different crystal planes295 and of the occurrence of atoms differing in their co-ordination number.141

The contribution that a surface atom makes to surface energy increases as its CN decreases. In a bimetallic particle, the atoms of the metal of lower surface energy (due to its lower heat of sublimation) will therefore be expected to congregate at low CN sites. Monte Carlo calculations10,296,297 based on pair-wise bond interactions leading to surface configurations that minimise the total energy have been carried out for a number of Groups 8–10 and Group 11 pairs (Pt-Cu, Rh-Ag etc.).10,298 They confirm that for cubo-octahedra it is edges and vertices that are first occupied by Group 11 atoms, followed by (100) planes; (111) planes are filled last. These observations may vitiate conclusions on the sizes of active centres deduced by altering the composition of an alloy, if it is assumed that the inactive atoms are randomly distributed over the surface.280 Considerations of size and of specific chemical bond formation299 may also determine the location of the inactive element.

The molecular orbital approach to describing free valencies at metal surfaces (Section 1.23) has been extended to treat small particles.175,178,300,301

2.5.6. Conclusions

The purpose of the last sections was to create an overview of the changes which occur in the physical properties of metal particles as their sizes alter, and the ways in which these may be explained, in order to provide a basis for understanding particle size effects in chemisorption and catalysis. The complications that may arise from the having a binodal distribution of particles size have however always to be kept in mind. The main conclusions have been summarised by Burch6 in the following way.

(1) Proportions of atoms having unusually low CN will change relatively little as size is increased beyond about 2 nm (ca. 60 % dispersion). (2) Bulk electronic properties are not likely to be shown by particles having less than about 150 atoms (1.7 nm, 70 % dispersion) (3) Typical surface electronic properties are however probably shown by particles with only 25–30 atoms (90 % dispersion). (4) Unusual crystallographic structures are rare. (5) Since the heat released by chemisorption is enough to convert one structure into another, metastable structures are unlikely to survive under reaction conditions, and surface reconstruction may occur frequently.

(6) A small metal particle may comprise a solid core and a semi-fluid surface layer.

To this catalogue may be added the worry (7) that the size distribution may be binodal.

It is usual but not necessarily sensible to regard geometric and electronic structures as quite separate things, whereas in reality they are closely connected. The form adopted by a metallic or bimetallic particle will depend on the bonds formed between the component atoms, that is to say, on its electronic structure; and, because of the mobility of surface atoms, the properties of very small particles (dispersion >50%) will be dominated by electronic factors.

2.6. METAL-SUPPORT INTERACTIONS

2.6.1. Causes and Mechanisms

The point has already been made that it is almost impossible to study metal particles unless they are supported in the same way. The extent of interaction between the metal and support may be classified as either weak or moderate or strong. With metal particles formed by condensation of atoms from the vapour or by chemical means onto supports that are neutral in character (e.g. graphite), the interaction will probably be weak and the particles will show normal intrinsic effects of size. Complications such as they are will be limited to pseudomorphism caused by epitaxial growth on crystalline supports.273 Larger effects occur with particles formed by chemical means on supports such as alumina, where the forces responsible for anchoring them to the support have often been discussed, and have now been characterised (see later). Even larger consequences for the metal arise if it finds itself in contact with or near to protons (e.g. in zeolites) or basic cations (e.g. in zeolite LTL or hydrotalcites), but the largest effects of all happen when the support is to some extent reduced by hydrogen: a phenomenon then occurs that has been given the name Strong Metal-Support Interaction (SMSI).91,199,243,244,302−305

This interaction is a consequence of the migration of entities from the support to the metal, with consequent blocking of the active surface. This happens with a number of reducible oxides, and is a result of hydrogen atoms moving from the metal to the support, i.e. of hydrogen spillover, which will be discussed in Chapter 3. Consideration of the striking consequences of the SMSI will also be deferred to that place.

The strength of the metal-support interaction is however a continuously varying commodity, depending in each system on the pre-treatment conditions, especially temperature. Since a metal-support interaction is most likely to be visible at small particle sizes and to diminish as size increases, it is natural to try to explore the effect by systematically altering the size: this of course means assuming the absence of intrinsic effects. There is however another problem: the methods commonly used to produce different particle sizes (alteration of metal loading, calcination, change of reduction temperature etc), may lead to other consequences

Figure 2.8. Contact angle and interfacial energies for a structureless metal particle on a support.

affecting the number or type of catalytically active sites, e.g. various concentrations of anions (e.g. Cl– which is strongly retained by alumina and titania), formation of toxic species from impurities in the support (e.g. SO42− which generates H2S during reduction) or, even with supports as hard to reduce as alumina, silica or magnesia, the partial obscuration of the metal through one of the several causes (see later). There are ways (some simple) of avoiding or at least recognising all these potential problems, although they are not always adopted. Intrusion of such phenomena can create a false impression of particle size and metal support effects. The intended change is therefore not always that which is produced, or not only that. The good that I would, I do not. . . .

For a structureless particle or liquid drop on a support, the contact angle θ that it makes at equilibrium depends upon the surface energies at the three interfaces (see Figure 2.8) according to the equation306

The corresponding forms adopted by crystalline particles as the interaction across the metal/support interface increases are also shown. In some systems, especially those subject to SMSI, particles are occasionally seen to form raft-like structures307 indicative of a very strong interaction, but for others where the interaction is less strong they are usually described as hemispheres, half-cubo-octahedra, or square pyramids that may be truncated.6,164 In the presence of an adsorbing gas, the particle shape will depend on the adsorption energy through the γmg term, and the differential energies at the various faces will determine the extents of their exposure; the effect on the γsg term will usually be negligible.

The structure at the metal-support interface is the critical factor in controlling particle shape and stability.295 A number of careful studies using EXAFS have illuminated it, and the changes that occur when reduction temperature is increased. Systems subjected to particular study are listed in Table 2.1. In addition to

Figure 2.9. Changes in the metal-oxygen (M. . . O2−) distances observed by EXAFS with Rh/Al2 O3 : the longer distance occurs when H atoms are present at the interface (right-hand part), and the shorter distance when they are removed by evacuation or oxidation (left-hand part). The sizes of the various components are only approximately to scale.

metal-metal distances, metal-oxygen distances have also been determined; these decrease at high reduction temperature or (in the case of alumina306,307) with outgassing, due it is thought to the irretrievable loss of hydrogen emanating from the surface hydroxyls (Figure 2.9).308 The closer contact between metal (or interfacial metal ions) and oxygen, coupled with the loss of protons, may constitute the ‘chemical glue’, which has long been suspected to account for the stability of supported metal particles. The same changes are often seen with very small palladium, platinum or ruthenium particles in acidic or basic zeolites,47,68−70,157,191,309,310 such particles frequently appearing to have an electron-deficient character.68,69,311

Several explanations have been suggested for these effects. (1) There may develop a degree of chemical bonding, covalent or more likely ionic, between the bottom of metal atoms and the oxide ions of the support; this might give these atoms some positive charge and make the whole particle electron-deficient (Figure 2.10A):

The existence of metal cations close to small metal particles in Ru/Al2O3 has been noted,112 but they are not seen in Rh/Al2O3.306 They are however commonly seen with palladium catalysts, and electron-deficient palladium and Pdx+ cations are assigned unusually high catalytic activity in hydrocarbon reactions218 (Figure 2.10B).

(2) Such species (and similar platinum ones212,213,215,312,313) could arise in acidic zeolites or on other acidic supports in the following way:

No desorption of hydrogen is needed for this (Figure 2.10C). (3) More recently the study of zeolite LTL and of silica with very small pallium or platinum particles, and various levels of alkalinity in the form of K+, has suggested yet another explanation.

This arises from analysis of peaks in the Fourier transform of EXAFS spectra at distances less than 0.15 nm, due to the scattering of the emergent photoelectron against the atomic potential of the absorber atom itself: the effect is named Atomic XAFS (AXAFS).65 It is noted that the observed changes in XPS binding energies, XANES shape resonance (see Chapter 3) and AXAFS with basicity cannot be due to a simple polarisation or redistribution of charge caused by a neighbouring cation and that, although charge transfer from oxide ions of the support would increase electron density on the metal, this is unlikely to occur, especially when the support is an insulator. The proposed model relies on a change in the energies of the metal valence electrons, with the metal’s ionisation potential decreasing with increasing alkalinity. The primary interaction is Coulombic interaction between metal and oxide ions, affecting the interatomic potential but avoiding the need for actual electron transfer (Figure 2.10D).314 Two questions remain. (i) Is this the basis of a general explanation of metal support interactions? Does it help to explain the different activities reported for platinum supported on alumina, silica and other ceramic and semi-conducting supports through a variable elecron density on the oxygen? (ii) How exactly do the suggested changes in electron energy levels affect catalytic activity? Increasing alkalinity seriously depresses activity of ruthenium315 and platinum65,316 for hydrogenloysis and skeletal isomerisation; this does not however agree with the high activity of electron-deficient palladium,218 suggesting that Pdx+ ions have an opposite effect to that caused by K+ ions.

To confuse the situation further, differences in the vibration frequencies of M–C bonds in chemisorbed carbon monoxide and M–H bonds in chemisorbed hydrogen, between those found with small platinum particles in NaX zeolite317 and in magnesium aluminium hydrotalcite,74 and those for larger particles or bulk metal, have been interpreted to imply that the metal is negatively charged. It is however difficult to see how a particle can be both negatively charged and electron deficient at one and at the same time.

Attempts to eliminate residual metal cations by very high temperature reduction318 are usually unsuccessful because of the harmful effects caused to the already reduced metal307 (Section 2.32). It has often been found that this treatment produces a loss of catalytic activity larger than can be explained by sintering,91,305,319,320 and that rates decrease faster than hydrogen chemisorption capacity, thus giving apparently larger values of TOF. These effects have been found with Pt/SiO2 (EUROPT–1),305 Pt/Al2O3,128,321,322 Pt/MgO323 and Pd/SiO2,218,324 as well as other systems, and there is good evidence for at least some formation of intermetallic compounds with silicon or aluminium, or other elements:325 partial or substantial coverage of the metal by the support is another

possibility. Reduction of support cations must be preceded by migration of hydrogen atoms from the metal by hydrogen spillover (Section 3.34). Where support reduction is easier we find the SMSI (see Section 3.3.5), so consideration of titania and similar supports is deferred until then. It has also been thought that metals can retain some very strongly held hydrogen after high temperature treatment, and that this acts as a poison (Section 2.3.2); the effects of forming palladium hydrides on catalytic activities is also well appreciated. These matters are considered again in Section 3.3.4 and elsewhere.

2.6.2.Particle Size Effects and Metal-Support Interactions: Summary

It is difficult if not impossible to separate these two effects completely because they are so closely interlocked.

Apparent size effects and interactions can be caused by poisons originating in the support or metal precursor, or can be a consequence of thermal treatment leading for example to the formation of intermetallic compounds through reduction of cations of the support.

Bifunctional and spillover catalysis and reaction at interfacial sites may occur to an extent depending upon the components of the catalyst.

Particle size, shape and relative areas of different crystal planes can be determined by the way in which metal precursor reacts with the support surface and by epitaxial effects of the support on the metal particle during reduction.

The number of metal particles formed, and hence their mean size, may be conditioned by the concentration of defect sites on the support surface that act as nucleating points.

Small metal particles are more liable to feel support effects than large ones.

Particle size effects can be caused by (i) consequences of structural epitaxy;

(ii)variation of electronic constitution due to occurrences at the metalsupport interface; (iii) alteration in the extent to which electron energy levels overlap and form bands, so that the particle acquires metallic character;

(iv)changes in the population of atoms of specified coordination number, or of ensembles of atoms having specified characteristics; (v) alteration in the mobility of surface atoms associated with the variation of melting temperature with size; (vi) a difference in the number or proportion of atoms having special character due their being at the periphery of the particle, and hence able to collaborate with the support in achieving the reaction.

Models for metal-support interactions occurring at moderate temperatures include (i) Mn+ cations adjacent to the metal particle (Figure 2.9A),

(ii)unreduced Mn+ cations by proton acquisition in acidic zeolites (Figure 2.9B); and (iii) formation of Mn+ cations by proton acquisition in acidic zeolites (Figure 2.9C): these all give the metal an electron-deficient